Biomedical Diagnosis of Breast Cancer Using Deep Learning and Multiple Classifiers

Abstract

Breast cancer is considered one of the deadliest diseases in women. Due to the risk and threat it poses, the world has agreed to hold a breast cancer awareness day in October, encouraging women to perform mammogram inspections. This inspection may prevent breast-cancer-related deaths or reduce the death rate. The identification and classification of breast cancer are challenging tasks. The most commonly known procedure of breast cancer detection is performed by using mammographic images. Recently implemented algorithms suffer from generating accuracy below expectations, and their computational complexity is high. To resolve these issues, this paper proposes a fully automated biomedical diagnosis system of breast cancer using an AlexNet, a type of Convolutional Neural Network (CNN), and multiple classifiers to identify and classify breast cancer. This system utilizes a neuro-fuzzy method, a segmentation algorithm, and various classifiers to reach a higher accuracy than other systems have achieved. Numerous features are extracted to detect and categorize breast cancer. Three datasets from Kaggle were tested to validate the proposed system. The performance evaluation is performed with quantitative and qualitative accuracy, precision, recall, specificity, and F-score. In addition, a comparative assessment is performed between the proposed system and some works of literature. This assessment shows that the presented algorithm provides better classification results and outperforms other systems in all parameters. Its average accuracy is over 98.6%, while other metrics are more than 98%. This research indicates that this approach can be applied to assist doctors in diagnosing breast cancer correctly.

Breast Cancer Detection using Mammography: Image Processing to Deep Learning

“…Another study introduced a fully automated system for diagnosing breast cancer, utilizing an AlexNet and multiple classifiers to attain heightened accuracy level. Validation through testing on three Kaggle datasets confirmed its superior performance, indicating its potential utility in aiding medical professionals with precise diagnosis [138]. Marathe et al [139] introduced a quantitative method to differentiate between benign and actionable amorphous calcifications on mammograms.…”

Benign and Malignant Breast Tumor Classification in Ultrasound and Mammography Images via Fusion of Deep Learning and Handcraft Features

“…The system proposed in [18] employed the BUSI dataset, where the authors, because of the imbalance between the classes, performed data augmentation via generating synthetic samples that could duplicate data. Analyzing the system of Alsheikhy [19], we can see that their dataset was unbalanced-there were 1778 Malignant, 1408 Benign, and 185 Healthy images. Performing the classification process in such a dataset favored classification for the first two classes.…”

“…Table 10 shows the performance comparison of the designed systems versus different proposals published in the literature, which also used binary classification [11][12][13][15][16][17]20,21]. Different systems used the same datasets or subsets thereof [12,17,20], demonstrating outstanding performance [11,16,19], although several proposals did not provide access to their private datasets.…”

“…Alsheikhy et al [19] presented a study that used the AlexNET CNN architecture, employing different classifiers such as K-Nearest Neighbor (KNN), Naive Bayes with the Gaussian kernel, and Decision Tree (DT). DWT was employed for the images in an attempt to denoise White Gaussian Noise.…”